Plant Soil (2017) 413:97–110 DOI 10.1007/s11104-016-3085-9

REGULAR ARTICLE

Seed selection by earthworms: chemical seed properties matter more than morphological traits Julia Clause & Estelle Forey & Nico Eisenhauer & Charlotte E. Seal & Anne Soudey & Louise Colville & Sébastien Barot

Received: 12 June 2016 / Accepted: 7 October 2016 / Published online: 18 October 2016 # Springer International Publishing Switzerland 2016

Abstract Aims The passage of seeds through the earthworm gut potentially damages seeds, altering seed and seedling performances depending on seed traits. This work was conducted to study to what extent chemical and morphological seed traits determine the seed attractiveness for earthworms. Methods We tested seed selection via the ingestion and digestion of 23 grassland plant species spanning a range of 14 morphological and chemical traits by two common earthworm species: the anecic Lumbricus terrestris and the endogeic Allolobophora chlorotica. Results Both earthworm species ingested seeds from all plant species. A. chlorotica digested almost all ingested

seeds (out of the 15 % ingested), whereas L. terrestris excreted them in varying quantities (out of the 86 % ingested), depending on plant species identity. Seed ingestion rate by L. terrestris was driven by seed oil content and earthworm initial weight. The apparent effect of seed length was explained via seed oil content. Seed digestion rate by L. terrestris was negatively impacted by seed size. Seed ingestion rate by A. chlorotica tended to be impacted by seed protein content and seed length. Conclusion Earthworms–seed interactions depend on a variety of seed traits and earthworm identity. Thus, earthworms, via their specific feeding behavior, might facilitate or impede the regeneration of certain plant species and drive plant communities.

Responsible Editor: Erik J. Joner. Electronic supplementary material The online version of this article (doi:10.1007/s11104-016-3085-9) contains supplementary material, which is available to authorized users. J. Clause (*) : E. Forey : A. Soudey Normandie Univ, UNIROUEN, IRSTEA, ECODIV, 76000 Rouen, France e-mail: [email protected]

C. E. Seal : L. Colville Department of Comparative Plant and Fungal Biology, Royal Botanic Gardens, Kew, Wakehurst Place, Ardingly, West Sussex RH17 6TN, UK S. Barot IRD - iEES Paris, 7, quai St Bernard, 75230 Paris cedex 05, France

N. Eisenhauer German Centre for Integrative Biodiversity Research (iDiv) Halle-Jena-Leipzig, 04103 Leipzig, Germany

N. Eisenhauer Institute of Biology, Leipzig University, 04103 Leipzig, Germany

Present Address: J. Clause Laboratoire Ecologie & Biologie des Interactions—UMR CNRS 7267, Equipe Ecologie Evolution Symbiose, Université de Poitiers, 5, rue Albert Turpain, TSA 51106, 86073 Poitiers Cedex 9, France

98

Keywords Above-belowground interactions . Fatty acid composition . Granivory . Oil content . Seed predation . Seed size

Introduction The impact of earthworms on plants through the modification of soil properties has been shown extensively (van Groenigen et al. 2014). Earthworm impacts on plant communities through the ingestion of seeds and consequences for seed survival are, however, less well studied, although they have been shown to be significant (Forey et al. 2011). Earthworms transport seeds upward or downward through the soil (Willems and Huijsmans 1994; Zaller and Saxler 2007). Seed burial might be an essential mechanism to escape aboveground seed predation or harsh environmental conditions (Traba et al. 2006), but might also deprive access to light for germination (Donath and Eckstein 2012). On the other hand, transporting seeds upwards might facilitate the emergence of buried seeds (Decaëns et al. 2003). The ingestion of seeds by earthworms alters seed germination and seedling establishment (Aira and Piearce 2009; Eisenhauer et al. 2009). These alterations are likely due to seed damage in the earthworm gut (Curry and Schmidt 2007). In general, a percentage of ingested seeds (ranging between 0 and 100 %) that is both seed-specific and earthworm-specific was not recovered after ingestion because seeds were too damaged or digested (McRill and Sagar 1973). While some seeds are not viable anymore after gut passage, some viable seeds are excreted in nutrient-rich earthworm casts and burrows (Zhang and Schrader 1993), which may positively or negatively impact the plant development of viable seeds (Laossi et al. 2010). Animal feces are commonly recognized as regeneration niches, especially in cattle dung (Gardener et al. 1993). Therefore, seed selection by earthworms represents a first step in modifying plant community composition. These direct seedearthworm interactions are known to be seed- and earthworm-specific (Eisenhauer et al. 2009; Clause et al. 2011). Significantly different plant compositions were found between seed banks in cast and in soil along a chalk grassland ecotone, which suggests that seed ingestion by earthworms plays a significant role in plant composition dynamics (Clause et al. 2016). Seed palatability, which is the intrinsic property of the food that is assessed without considering post-

Plant Soil (2017) 413:97–110

ingestive consequences (Yeomans 1998), is likely determined by seed chemical and morphological properties for earthworms. Most studies show that earthworms mostly ingest seeds according to seed size (see Forey et al. 2011). Additionally, the effects of morphological seed traits, such as shape, volume, and mass, have rarely been tested (Grant 1983; Clause et al. 2011), and chemical traits of seeds have only been studied in one case (Clause et al. 2011). This last study indicated that the anecic earthworm Lumbricus terrestris and the epigeic Satchellius mammalis ingest seeds of chalk grasslands according to seed oil content as well as their size. This suggests that seed ingestion might be an ‘active’ ingestion process for nutritive purposes, i.e. granivory. Other types of chemical compounds might attract earthworms, such as protein content (Harrison et al. 2003) and volatile compounds (Paczkowski et al. 2013). Seed chemical and morphological traits are also likely to impact seed digestion by earthworms. Small seeds are generally digested or destroyed in higher proportions than large seeds because of contraction of the earthworm gizzard, grinding and enzymatic activity in the earthworm gut (Marhan and Scheu 2005; Curry and Schmidt 2007), and a longer retention time (Levey and Grajal 1991; Stanley and Lill 2002). Seed protein content and lipid content probably also influence seed digestion and nutrient uptake by earthworms (Clause et al. 2011). However, chemical and morphological seed traits may be correlated, and there has been no study on the relative importance of a broad spectrum of different seed traits for ingestion and digestion by earthworms thus far. Anecic and endogeic earthworm species are known to have different feeding behaviors. Anecic species mostly feed on plant litter, whereas endogeic species mostly feed on soil organic matter (Curry and Schmidt 2007). Both ingest seeds, although anecic L. terrestris ingests more seeds than endogeic species (Eisenhauer et al. 2009). Their feeding behaviors are associated with their anatomical external and internal characteristics, such as the size of their prostomium or the development of certain organs such as the typhlosolis, a mid-dorsal invagination in the midgut that may be involved in nutrient uptake efficiency (Makeschin 1997). Those species-specific differences might have an impact on seed ingestion or digestion. The objective of this study was to disentangle the influence of chemical traits (e.g. oil content) and morphological characteristics (e.g. seed size) on ingestion and digestion by earthworms. Therefore, we ran an

Plant Soil (2017) 413:97–110

experiment similar to that of Clause et al. (2011) with two earthworm species (one anecic and one endogeic species) and 23 seed species with very diverse chemical and morphological traits, including seed size and oil, protein, and fatty acid contents. We hypothesized that (i) the species compositions of ingested and digested seeds differ between L. terrestris and A. chlorotica, and (ii) their ingestion preferences depend on specific seed traits, most likely driven by seed size as well as oil and protein content.

Materials and methods Earthworm and seed species Ingestion of seeds by adult Lumbricus terrestris (4.28 ± 1.03 g, n = 230) and Allolobophora chlorotica (0.24 ± 0.05 g, n = 230) was tested. L. terrestris was commercially purchased and A. chlorotica was sampled from a wood border on chalk substrate in Hénouville (Upper-Normandy, France). All earthworms were kept in boxes containing soil and grass litter (Brachypodium perenne L., Lolium perenne L.) collected from a chalk grassland at least two weeks prior to the experiment (Fründ et al. 2010). Time between sampling and the experiment was between two and four weeks. All live earthworms were weighed with guts that were voided before the experimentation to test for an effect of earthworm weight on ingestion and digestion patterns. Seeds of 23 chalk grassland species were chosen for their wide range of size, mass, shape, texture, oil and protein contents across many plant families: Achillea millefolium L.; Agrostis capillaris L., Brachypodium pinnatum L., Carex flacca L., Centaurea nigra L., Daucus carota L., Deschampia cespitosa L., Festuca lemanii L., Galium mollugo L., Genista tinctoria L., Holcus lanatus L., Lolium perenne L., Lotus corniculatus L. Medicago lupulina L., Ononi spinosa L., Origanum vulgare L., Poa pratensis L., Ranunculus acris L., Sanguisorba minor L., Seseli libanotis L., Teucrium scorodonia L., Trifolium repens L., and Urtica dioica L. (see Online Resource 1 for details on seed traits). Seeds were commercially purchased from Emorgate Seeds (King’s Lynn, UK) or provided by the Caen and Bailleul Botanical gardens (France). ‘Seeds’ refers both to seeds and to fruits with not easily detached structures. For example, grass caryopses (except for B. pinnatum) and the fruit of S. minor were measured

99

as seeds. Microscopic control showed that seed processing prior to the experiment did not damage seeds. Germination rates of seeds prior to the experiment ranged from 0 % for C. nigra and G. tinctoria to 100 % for Origanum vulgare. Ingestion and digestion of seeds We followed the method of Eisenhauer et al. (2009) to study the ingestion, digestion, and excretion of seeds. After voiding their gut by making them fast for 48 h on moist filter paper (15 °C, darkness), earthworms were placed on moist filter paper in petri dishes (15 °C, 24 h, darkness) with 1 g of sieved soil (5 mm; collected from chalk grassland, Upper Normandy, France) with 20 seeds of a single plant species, placed at the soil surface. Thereafter, earthworms were removed and transferred into a different set of petri dishes with moist filter paper for 48 h (15 °C, darkness) to recover as many seeds as possible. Adding soil simulates natural conditions at the soil surface and provides sand particles that improve grinding and nutrient assimilation from organic matter in the earthworm gut (Marhan and Scheu 2005; Curry and Schmidt 2007). Each treatment combination (two earthworm treatments and 23 seed treatments) was replicated ten times (460 petri dishes in total). At the end of the experiment, casts were gently manually broken apart with water, and the number of non-ingested and egested seeds was counted. Seeds that were not recovered in casts were considered as non-ingested seeds, and egested seeds in casts were considered as ingested. The difference between the total number of seeds used (20) and the sum of the non-ingested and egested seeds was considered as the number of digested seeds. Digestion of seeds was only calculated for treatments where at least three earthworm individuals per species ingested at least three plant seeds, for statistical purposes. Seed traits selection and measurements Seed mass, size, and shape Fourteen traits were selected to test the impact of seed traits on ingestion and digestion by earthworms. For each plant species, seed mass was calculated as the average of 20 seeds, individually weighed with a precision balance (precision: 10−7 g). Seed length, width, and thickness were measured on 10 seeds with a Zeiss

100

AxioCam HR camera. Length ranged from 1.81 mm (O. vulgare) to 8.41 mm (S. minor) (Online Resource 1). Seed shape was estimated using Thompson et al. (1993) shape index on the 10 seeds, where the shape (Vs) is determined by dividing the seed length, width, and thickness by the length and by calculating the variance of these three values with Vs = [Σ (x – x¯)2/n]1/2 (Thompson et al. 1993). As such, values for seed shape vary between 0 (perfectly spherical) and 0.2 (elongated or disc-shaped seeds). Values ranged from 0.002 (T. repens) to 0.168 (F. lemanii) (Online Resource 1). Seed carbon and protein contents Seeds of each species were ground to obtain between 2.2 mg (O. vulgare) and 3.0 mg (L. corniculatus) of seed powder that were weighed and analyzed for C and N contents. The total carbon and nitrogen contents were measured by combustion on two replicates per seed species (CHN Analyzer, Fisher Scientific, Germany). Protein content for each sample was determined with the Kjehldal method (N × 6.25). Values for the total carbon content ranged from 42.4 % (L. perenne) to 54.2 % (O. vulgare) (Online Resource 1). Values for protein content varied from 14.3 % (L. perenne) to 41.8 % (O. spinosa) (Online Resource 1). Seed water content, oil content, and fatty acid composition Seed water content was calculated by subtracting the fresh weight of 10 seeds from the dry weight (107 °C, 17 h) and by dividing the result by the fresh weight. The mean value was calculated from five replicates. Values for the seed water content varied from 4.3 % (A. millefolium) to 19.4 % (C. flacca) (Online Resource 1). Oil content was extracted and quantified by supercritical fluid extraction with carbon dioxide according to Seal et al. (2008). For this, 0.5 g (± 0.001 g) of seeds were ground and mixed with 1.5 (± 0.001 g) of Wetsupport™ before further analysis with a ISCO SFX 3560 fat analyser (6000 psi, 80 °C). Oil content was quantified by weighing the final vacuum-dried (70 °C, 1 h, n = 3) extract and by dividing it by the initial seed dry weight in the mix. Controls were performed with 5 drops of sunflower oil. Values for oil content varied from 0.9 % (L. perenne) to 34.9 % (U. dioica) (Online Resource 1).

Plant Soil (2017) 413:97–110

Fatty acid composition of seeds was analyzed with a Gas Chromatography-Mass Spectrometry (GC-MS) according to Colville et al. (2012) after oil extraction. The compounds were detected using a Mass Spectrometer (Thermo Finnigan Trace DSQ; ionization energy 70 eV, scan frequency range m/z 10–500 per 0.3 s) and identified through comparison with the NIST mass spectral database and analytical standards (F.A.M.E. Mix C4C24, Supelco). Excalibur® software (Fisher Scientific) was used to facilitate the identification of the most abundant chemicals components. Quantification of fatty acid methyl esters was performed using standard curves of quantitative standard mixtures (F.A.M.E. Mix GLC10, −30 and −50, Supelco). Many fatty acids were detected in each seed species (Online Resource 2). In further analyses, we only kept the five fatty acids that were identified in all species and in the largest amounts: palmitic acid (16:0), stearic acid (18:0), oleic acid (18:1ω9), linoleic acid (18:2ω6), and α-linolenic acid (18:3ω3) (Online Resource 1). Data analysis To test the effects of seed species, earthworm species, and their interaction on ingestion and digestion, we ran analyses of variance using generalized linear models (GLM), with binomial or quasibinomial distribution to overcome overdispersion (see Zuur et al. 2007). Post hoc comparisons among seed species were performed within each earthworm species with a Tukey HSD test (α = 0.05). Means of ingested and digested seeds (%) were compared for each plant species between both earthworm species, with a 2-sample non-parametric Wilcoxon-Mann-Whitney test (α = 0.05). Due to the low ingestion by A. chlorotica, 13 plant species could not be considered in the comparison of digestion patterns between both earthworm species. The variability of seed morphological characteristics was described using a Principal Component Analysis (PCA) on a 12 traits × 23 seed species matrix. Variables were log-transformed to improve normality. Variables were then centered and standardized by standard deviation. Seed ingestion and digestion by L. terrestris and A. chlorotica, as well as earthworm weight at the start of the experiment, were added to the analysis as illustrative variables. Illustrative variables do not contribute to the correlation circle and to axes, but their correlation with PCA axes can be tested. Illustrative variables were projected on the PCA axes to see how they were

Plant Soil (2017) 413:97–110

associated with the different seed traits. Additional correlation tests were run between ingestion as well as digestion and each illustrative variable. To disentangle which seed traits determined ingestion and digestion by earthworms, while accounting for correlations among traits, we performed a structural equation modeling (SEM). The first step in SEM requires establishing an a priori model based on known and expected relationships among variables. Based on results of the PCA and regression analyses, we decided to only keep variables that showed significant or marginally significant correlations with response variables. Data followed the same transformations as for the PCA analysis, with an additional arcsine-transformation of continuous percentages of ingestion and digestion. We fitted the model by testing for the overall goodness of fit using the maximum likelihood (ML) estimation procedure. Relationships among the remaining variables were chosen based on our knowledge of seeds, and on the model fitting procedure to reach the best model as possible. Adequate model fits are indicated by a nonsignificant χ2 test (P > 0.05), low AIC and low Root Mean Square Error of Approximation (RMSEA) (Grace 2006). Results were interpreted by using standardized path coefficients (SPC) of the model and P values. Path coefficients are analogous to partial correlation coefficients. They describe the direction and the strength of a relationship between two variables. PCA analysis was carried out using the ade4 (Dray and Dufour 2007) and FactoMineR (Husson et al. 2013) modules within the R environment (R Core Team 2013). SEM was performed using Amos 5 (Amos Development Corporation, Crawfordville, FL, USA).

Results Ingestion of seeds Both earthworm species ingested seeds of all plant species. Seed ingestion rates depended on the interaction between the seed species and the earthworm species (GLM, χ2(45,412) = 6612, P < 0.001). Overall, the total ingestion rate was higher for L. terrestris (86 ± 1 %) than for A. chlorotica (15 ± 1 %; GLM, χ2(1456) = 5180, P < 0.001; Fig. 1). Higher seed ingestion for L. terrestris than for A. chlorotica was also found within each seed species, except for seeds of T. repens (Fig. 1 and Table 1).

101

Seed ingestion by L. terrestris varied among species (GLM, χ2(22,205) = 901, P < 0.001). L. terrestris ingested 100 % of seeds of O. vulgare, Urtica dioica, T. scorodonia, S. libanotis, and O. spinosa (Fig. 1). Ingestion rates were high (>75 %) for all other seed species except for seeds of F. lemanii (67 ± 6 %), B. pinnatum (52 ± 8 %), and T. repens (38 ± 9 %; Fig. 1). Seed ingestion by A. chlorotica also varied among species (GLM, χ2(22,207) = 530, p < 0.001). A. chlorotica ingested more seeds of O. spinosa (56 ± 3 %) and T. repens (49 ± 9 %) than of any other seed species (Fig. 1 and Table 1). The other seed species with similar ingestion rates were A. millefolium (28 ± 3 %), U. dioica (23 ± 5 %), T. scorodonia, L. perenne (18 ± 3 %), L. corniculatus (16 ± 2 %), D. carota (15 ± 4 %), and A. capillaris (13 ± 4 %). Digestion of seeds Seed digestion depended on the interaction between the seed species and the earthworm species (GLM, χ2(32, 288) = 1028, P < 0.001). Overall, A. chlorotica digested more of the ingested seeds (99 ± 1 %) than L. terrestris (45 ± 2 %; GLM, χ 2(1319) = 683, P < 0.001; Fig. 2 and Table 1). Seed digestion by L. terrestris varied among plant species (GLM, χ2(22,203) = 318, P < 0.001). The seeds with the highest digestion rates were that of T. repens (83 ± 7 %) and A. capillaris (73 ± 5 %), followed by D. carota (60 ± 5 %), A. millefolium (59 ± 5 %), G. mollugo (59 ± 7 %), B. pinnatum (57 ± 10 %), O. vulgare (54 ± 6 %), L. corniculatus (52 ± 9 %), and U. dioica (47 ± 6 %; Fig. 2 and Table 1). Seed digestion by A. chlorotica reached 100 % for all ingested seeds, except for A. capillaris that was less digested (GLM, χ2(9,85) = 28, P = 0.001; Fig. 2 and Table 1). Digestion of 13 seed species could not be determined due to low ingestion rates by A. chlorotica: grasses H. lanatus, D. cespitosa, C. flacca, P. pratensis, F. lemanii, B. pinnatum, herbs G. mollugo, C. nigra, R. acris, S. libanotis, and S. minor, and legumes M. lupulina and G. tinctoria. Correlative analyses of seed traits The first PCA axis explained ca. 28 % of the total variability between seed traits and was mostly associated with seed chemical traits: seed content in oil, carbon, and stearic, oleic, palmitic, and linoleic acids (Table 2 and

Plant Soil (2017) 413:97–110 100

102 a a de de a

ab

a ab a bc

a be

a bc

ad

a de

a a de be

a de a a bc bc

80

bd cd

df

A

60

AB ef

40

Ingestion (%)

a bc a bc

a bc

BC CD

20

E

E

DE E

ED

E

E

CE

CE DE E

DE

ED DE

E

n ce s C fla Pp ra Fl em Lp er Bp in n O vu U di Am il G l m ol Ts l co r D ca C r ni gr R a ac ris Sl ib Sm in o Tr r ep Lc or M lu p G tin c O t sp in D

H

Ac

ap

*** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** ns *** *** *** *** la

0

CE

CE

CE CE

Fig. 1 Number of seeds (% ± SEM) of 23 different seed species ingested by L. terrestris (grey) and A. chlorotica (black). Bars with different letters vary significantly within L. terrestris (lowercase) or within A. chlorotica (uppercase). Asterisks below bars indicate significant differences between ingestion patterns by both earthworms for each plant species (Wilcoxon-Mann-Whitney; *** P < 0.001, ns non-significant). Seed species were arranged according to plant functional identity (grasses, herbs, and legumes) and seed size (small-seeded species and large-seeded species):

A. capillaris (Acap), H. lanatus (Hlan), D. cespitosa (Dces), C. flacca (Cfla), P. pratensis (Ppra), F. lemanii (Flem), L. perenne (Lper), B. pinnatum (Bpinn), O. vulgare (Ovu), U. dioica (Udi), A. millefolium (Amill), G. mollugo (Gmoll), T. scorodonia (Tscor), D. carota (Dcar), C. nigra (Cnigra), R. acris (Racris), S. libanotis (Slib), S. minor (Sminor), T. repens (Trep), L. corniculatus (Lcor), M. lupulina (Mlup), G. tinctoria (Gtinct), O. spinosa (Ospin)

Fig. 3b). Six out of eight traits explaining axis 1 were chemical traits (Table 2). However, length and shape also significantly explained axis 1 (Table 2 and Fig. 3b). Long seeds that were enriched in lineoleic, oleic, stearic, and palmitic acids, and that were represented by grasses L. perenne, P. pratensis, D. cespitosa and B. pinnatum, were opposed to small, round, oil- and carbon-rich seeds, represented by U. dioica, O. vulgare, A. millefolium, and T. scorodonia (Figs 3a and b). Variability of ingestion and digestion rates, represented as illustrative variables, were not explained by this axis (Table 2). However, a high mean ingestion was associated with high values of protein, carbon, and oil content with α = 0.01. PCA axis 2 accounted for additional ca. 23 % of the total variability among seed traits, and appeared to be a size-related axis (Table 2 and Fig. 3b). It opposed wide, thick, and heavy seeds represented by O. spinosa, S. libanotis, R. acris, and G. tinctoria, to narrow, shallow, or light seeds, such as A. capillaris, D. cespitosa, O. vulgare, P. pratensis, A. millefolium, and F. lemanii (Figs 3a and b). Digestion of seeds by L. terrestris was significantly represented by this axis (Table 2).

L. terrestris digested preferentially small, round, light seeds in contrast to wide, thick, and heavy ones. Axis 3 accounted for ca. 14 % of the total variability among seed traits. As for axis 1, it was strongly associated with seed fatty acid content, here alpha-linolenic, palmitic, and stearic acids (Table 2). It was also strongly associated with seed protein content (Table 2), and opposed G. tinctoria and O. vulgare to D. carota, F. lemanii, C. flacca, B. pinnatum, and S. libanotis (data not shown). Individual correlations showed that mean seed ingestion and ingestion of seeds by L. terrestris were significantly positively correlated with seed oil and carbon contents, and marginally negatively correlated with seed length (Table 3). Seed digestion by L. terrestris was significantly negatively correlated to seed size, i.e. length, width, thickness, volume and mass, and marginally negatively correlated with stearic acid content. Seed ingestion by A. chlorotica was not significantly correlated to any seed trait, but tended to be positively correlated with seed protein content, and to be negatively correlated with seed length (Table 3).

Plant Soil (2017) 413:97–110

103

Table 1 Summary of the generalized linear model (GLM; binomial) to test the impact of earthworm species on seed ingestion and digestion (n = 10 microcosms). Df: degree of freedom, residual degree of freedom Seed species

Ingested seeds

Digested seeds

df

χ2

P

df

χ2

P

Total

1456

5180